Text-guided CLIP-DINOv2 similarity maps, when added as prompts to SAM-HQ, improve segmentation of thin and complex objects on BIG, ThinObject5K, and DIS5K.
In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2023)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Talk2SAM: Text-Guided Semantic Enhancement for Complex-Shaped Object Segmentation
Text-guided CLIP-DINOv2 similarity maps, when added as prompts to SAM-HQ, improve segmentation of thin and complex objects on BIG, ThinObject5K, and DIS5K.